Time-Series in Structure-from-Motion Photogrammetry: Testing Co-Registration Approaches for Topographic Change Analysis

In the present work, we show preliminary results on testing co-registration techniques for topographic data derived from Structure-from-Motion photogrammetry, based on a multi-epoch photo alignment. The underlying goal is to enable the co-registration of large datasets (i.e., hundreds to thousands of photographs) by avoiding recalculating the entire time-series when a new epoch is added to the data collection. Tests are performed with data acquired in a harsh volcanic environment, where ideal and similar datasets are impossible to acquire. Although additional investigation is required to get an overall picture of the best co-registration methods to follow, these preliminary results highlight that, if multi-epoch co-alignment works well for homogeneous time-series (i.e., same camera and acquisition geometry), it is actually more complex for less ideal data, where the best methodology to follow depends on the ground control points available and the heterogeneity of the datasets involved in the multi-epoch co-alignment process.